[sdiy] Where to buy ceramic caps

[David G Dixon]

Well, I see that the passive-aggressive behaviour has migrated from Muff
Wiggler to the SDIY forum.  That's too bad.

As far as ego-wang, I don't need to be lectured on how to design a circuit
-- I've done a lot of it, quite successfully, thanks.  That's all I'm
saying.

Also, I never asked anyone to help me design anything.  This MAY be an
Intellijel product (and I find it a bit dismaying to see how much animosity
that seems to generate here) -- that depends on whether or not I can
convince Danjel to manufacture a purely analog frequency shifter in 2018,
when everything else out there is digital and cheap as chips to build.  I've
already designed the damn thing, built it (on a slab of plywood) and tested
it.  I just asked where (if) I could get some 1% leaded capacitors.  

It's not design I have a problem with -- it's procurement.  I hate buying
this stuff.  Having to buy this shit from dealers like Digikey and Mouser
almost makes me want to find a different hobby.  Going on those websites
with tens of thousands of options makes me want to puke, and it's why I
spend so much time and money at my local shop, Lee's Electronics in
Vancouver.  The only online retailer I really like is Small Bear, because
there is a drag-down menu for component values.  On Digikey, the same
capacitor in different values is spread across dozens of pages, each with a
dizzying array of options, most of which I don't understand.

And I have considered how sensitive the circuit is to precision.  That's
what all the simulation is for.  It turns out that, as long as both the
capacitors and the resistors  fall within 1% tolerance, it's almost as good
as using perfect parts (phase shift well within 1 degree of 90 across the
audio range -- good enough for me).

Interestingly, I found that I had 23 gold (probably Kemet) 103J caps (10nF,
5% tolerance) in an envelope, so I measured them.  Here are their values, in
ascending order in nF:

9.66
9.68
9.71
9.76
9.78
9.82
9.87
9.87
9.88
9.90
9.91
9.92
9.93
9.93
9.94
9.96
10.00
10.00
10.08
10.11
10.12
10.16
10.20

This works out to 9.92 +/- 0.15.  So, the actual tolerance is about 1.5%,
but the mean capacitance is 0.08% less than the stated value.  I put the 12
closest caps (9.87n to 10.00n) in my simulation, in random order, with 1%
resistor tolerances, and got almost perfect phase displacement.  This will
work for handbuilt units -- it took me about 2 minutes to test all those
caps.  It wasn't so good when I put every other cap in the list into the
simulation -- error was as much as 2 or 3 degrees.  Hence, selection will be
important unless 1% caps can be obtained.

I'm not too worried about the 3M resistor.  Johnson noise is proportional to
the square root of resistance, so 3M generates 10 times as much noise as
30k.  Given that the noise is very small to start with, I think I can live
with that.

Also, I'm using JFET-input opamps, so I'm also not too worried about the
different input impedances.  If I notice a significant difference between
this unit and the one I already built with all different caps and trimmers,
then I'll modify it.  I doubt I will.  Usually I find that all the stuff
people worry doesn't amount to much -- it's usually other unforeseen stuff
that bites one in the hindquarters.  The thing is that, if I try to keep the
resistors below 1M, then I'll have to use a 33nF cap in that one stage, and
33nF is well above the cutoff for C0G and 5% tolerance, so things get a bit
dicey.  I guess if it's just the one cap, I can hand-select for my own
builds, but that's not feasible with SMD parts.  I guess I'd rather have a
tiny bit of noise and a shifter that works properly and is easy to build.


> -----Original Message-----
> From: Neil Johnson [mailto:neil.johnson71 at gmail.com] 
> Sent: Saturday, December 30, 2017 2:15 PM
> To: David G Dixon
> Cc: Michael E Caloroso; SDIY
> Subject: Re: [sdiy] Where to buy ceramic caps
> 
> Hi,
> 
> David G Dixon wrote:
> > I always build actual circuits, but I find that my simulations are 
> > usually perfect.  Several million dollars worth of my 
> module designs 
> > have been sold to eurorack users around the world, so I 
> think I know 
> > how to design a circuit that works.
> 
> Careful David, your "ego-wang" is showing.
> 
> > Well, my design is about as easy as it gets: Two parallel 
> sets of six 
> > active all-pass stages in series, each stage tuned to a certain 
> > frequency as predicted by the awesome QuadNet program.  And 
> it works, very very well.
> > However, the RC values have to be fairly precise to get really 
> > superior performance.
> 
> With precision you then need to consider how sensitive the 
> circuit is to changes in the component values (from 
> tolerance, from soldering, from age, etc).
> 
> > This stupid thing is going to be a eurorack module in the 
> > not-too-distant future, and all the parts will be SMD.  I 
> presume that 
> > SMD multilayer caps can be easily and cheaply acquired in any 
> > tolerance you like, but I don't use SMD components in my 
> own personal 
> > builds, because I cant see them, and I drop them, and I wanna kill 
> > myself whenever I have to solder an SMD part by hand.  I'd 
> rather just not do electronics than work with SMD parts by hand.
> 
> Oh, right, so we're designing the next great product for 
> Intellijel.  Cool!
> 
> In which case you're painting yourself into a corner right 
> from the start since production will be using very different 
> parts, and if you're not careful you'll (sorry, WE'll) be 
> designing a product that is too expensive to manufacture.  
> There's a whole world of difference between making a one-off 
> prototype and designing something that needs to go into 
> production using a very different assembly process.
> 
> Neil
> --
> http://www.njohnson.co.uk